Peer groups splitting in Croatian EQA scheme : a trade-off between homogeneity and sample size number (CROSBI ID 237684)
Prilog u časopisu | izvorni znanstveni rad | međunarodna recenzija
Podaci o odgovornosti
Vlašić Tanasković, Jelena ; Coucke, Wim ; Leniček Krleža, Jasna ; Vuković Rodriguez, Jadranka
engleski
Peer groups splitting in Croatian EQA scheme : a trade-off between homogeneity and sample size number
Laboratory evaluation through external quality assessment (EQA) schemes is often performed as ‘peer group’ comparison under the assumption that matrix effects influence the comparisons between results of different methods, for analytes where no commutable materials with reference value assignment are available. With EQA schemes that are not large but have many available instruments and reagent options for same analyte, homogenous peer groups must be created with adequate number of results to enable satisfactory statistical evaluation. We proposed a multivariate analysis of variance (MANOVA)-based test to evaluate heterogeneity of peer groups within the Croatian EQA biochemistry scheme and identify groups where further splitting might improve laboratory evaluation. EQA biochemistry results were divided according to instruments used per analyte and the MANOVA test was used to verify statistically significant differences between subgroups. The number of samples was determined by sample size calculation ensuring a power of 90% and allowing the false flagging rate to increase not more than 5%. When statistically significant differences between subgroups were found, clear improvement of laboratory evaluation was assessed before splitting groups. After evaluating 29 peer groups, we found strong evidence for further splitting of six groups. Overall improvement of 6% reported results were observed, with the percentage being as high as 27.4% for one particular method. Defining maximal allowable differences between subgroups based on flagging rate change, followed by sample size planning and MANOVA, identifies heterogeneous peer groups where further splitting improves laboratory evaluation and enables continuous monitoring for peer group heterogeneity within EQA schemes.
external quality assessment ; multivariate analysis of variance ; peer group
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o izdanju
55 (4)
2017.
539-545
objavljeno
1434-6621
10.1515/cclm-2016-0284